Insights / The Future of Medical-Legal Services
Why the Future of Medical Record Review Belongs to Firms That Combine AI With Human Expertise
By Breen LNC.
Short answer: The work has two parts, and each suits a different worker. Software brings speed across thousands of pages. A nurse brings judgment, context and accountability. The hospital evidence so far suggests AI drafts with human review may be viable, which is why we think the future belongs to firms that do both.
Doing it by hand is getting harder
Records keep growing. The median number of notes available at an emergency visit rose from 5 to 359 between 2006 and 2022 at one health system. Outpatient notes grew by 60.1% in median length over a decade at another. These numbers aren’t litigation files, but they’re where litigation files come from. Reading it all by hand is a harder job every year.
AI alone isn’t the answer
Published evidence shows why. AI hospital narratives had more errors per summary and were less comprehensive than physician narratives. In a 2026 pilot, physicians reported omissions in a quarter of rated summaries. NIST warns of confident false output and of over-reliance by the people reading it. And in court, Mata v. Avianca showed where unchecked output can lead.
Where the combination has evidence behind it
Three hospital studies point the same way. The JAMA Internal Medicine authors suggested AI narratives with human review may be viable. The npj Digital Medicine authors, who work at a company building these tools, reported that refining prompts and workflows reduced major errors, in their own setup. The same pilot that reported omissions also reported that most unedited summaries carried no potential for harm, and that burnout fell. Those results come from hospitals, and we’d love to see the same work done on litigation chronologies. So far we haven’t found any.
Professional bodies say much the same. The American Nurses Association says AI “augments, supports, and streamlines expert clinical practice.” Arizona’s guidance for lawyers says to check the work product and that “legal professionals must not defer to its judgment.”
What combining them looks like
It’s more than adding a person at the end. In a working system:
- Software reads all the pages, marks repeats and scans, and drafts entries with a page cite.
- A nurse decides each entry with the page beside it, and adds what’s missing.
- A second reviewer checks the work.
- The attorney receives what the record shows, what needs attention and what’s still outstanding.
- Everything can be traced to a page and a person.
What firms will do differently
We expect firms that adopt this to spend less time looking for facts and more deciding what to do with them. They’ll ask new questions of vendors and of themselves. Some will build it, and many will partner for it. That’s a prediction, and it could be wrong in the details.
The principle seems safe to us. Attorneys need the record read quickly, and they need it read correctly. Those used to be in tension. Our whole position comes down to one line: the speed of AI, the judgment of a nurse.
Frequently asked questions
Is AI or human medical record review better?
The evidence we reviewed comes from hospitals, not litigation, and it suggests AI drafts with human review may be viable. We did not find a direct study of litigation chronologies.
What does combining AI with human expertise look like?
Software reads and drafts with page cites. A nurse decides each entry against its page and looks for what is missing. A second reviewer checks the work, and the attorney decides what to do with it.
Sources
- Patterson et al., Call me Dr Ishmael: trends in electronic health record notes available at emergency department visits and admissions, JAMIA Open (2024)
- Rule et al., Length and Redundancy of Outpatient Progress Notes Across a Decade at an Academic Medical Center, JAMA Network Open (2021)
- Williams et al., Physician- and Large Language Model-Generated Hospital Discharge Summaries, JAMA Internal Medicine (2025)
- Grolleau et al., Physician-Reported Safety Outcomes of AI-Generated Hospital Course Summaries, JAMA Network Open (2026)
- NIST AI 600-1, AI Risk Management Framework: Generative AI Profile (July 2024)
- Mata v. Avianca, Inc., No. 22-cv-1461, Opinion and Order on Sanctions (S.D.N.Y. June 22, 2023)
- Asgari et al., A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation, npj Digital Medicine (2025)
- American Nurses Association, The Ethical Use of Artificial Intelligence in Nursing Practice, OJIN Vol. 30, No. 2 (May 2025)
- Supreme Court of Arizona, Steering Committee on AI and the Courts, Generative AI: Ethical Best Practices for Lawyers and Judges (Nov. 14, 2024)
General information for legal professionals. This is not legal or medical advice.